Alright. so i’m comparing two GPUs and see one is 322% faster than the other.
And if my thought process is correct here:
100% is double, that’s simple enough. 200% would be double of the initial DOUBLE which would be 4x? Plus another 100% …? A total of EIGHT times the speed?
In case anyone is wondering i’m comparing the 960 to my old HD 4850
I suppose it’s correct? I read elsewhere that it would be quadruple. But that doesn’t make much sense to me. It progresses in a linear-ish way. 100% means you’re doubling whatever is the current reference point. If you’re dealing with consecutive 100s then they’re basically exponents, are they not?
I guess it adds up. Kind of crazy to see that sort of comparison, but the 4850 is an 11 year old card now. And by comparison, 7 years older than the 960. So I suppose it makes sense.
Usually when benchmark sites compare two products, the slower one is arbitrarily set as the baseline (100%) and the faster one is described with respect to that.
322% means the faster GPU has 322% of the performance of the slower one. It is not 322% faster than the slower one. It is actually only (322-100=) 222% faster than the slower one.
322% is 3.22x “as fast as” the slower one, or 2.22x “faster than” the slower one.
To see how fast one is compared to the other, divide the faster number by the slower number.
To see how much faster one is than the other, subtract the slower number from the faster number.
If 100% is the baseline, then 200% is double (2x), 400% is quadruple (4x) and 800% is octuple (8x). To double the speed the first time, you need to go from 100 to 200% — so that’s an increase of 100% of the base amount. But to double it again you need to go from 200 to 400% — so that’s an increase of 200% of the base amount. To double it a third time you need to get from 400 to 800% — an increase of 400% of the base amount. Every time you want to double something, you need to look at the current number and get to a number that is 2x that. Simply adding 100% of the base only works the very first time you want to double something — it doesn’t work after that
Doubling is an exponential (technically ‘geometric’) process, not a linear (technically ‘arithmetic’) one — so you do it by multiplying, not by adding.
Ok, makes sense now. If that’s the case then i’m actually overwhelmed. Or underwhelmed depending on how you look at it. So the 960 is actually 2.22x faster than the old 4850. WOW… Hmm?
I thought it would be much quicker.
And when comparing the 4850 to the 970, WOAH! It’s 566% faster, so 4.66x faster?! SHEESH!
“The 970 is 566% as fast as a 4850” or “The 970 is 566% the speed of a 4850”
“The 970 is 466% or 4.66x faster than a 4850”
Use the word “faster” only when you’re talking about what goes beyond 100% (or 1x).
Use something like “as fast as” or “the speed of” in all other situations.
